Accumulation of mutant p53 protein occurs frequently in human malignancies, including 40-60% of non-small cell lung carcinomas. The implications of such p53 over-expression, usually assessed by immunohistochemical techniques, for the prognosis of lung cancer patients remain undetermined. In this study, we used a time-resolved immunofluorometric assay to measure p53 protein concentrations in extracts prepared from 86 primary non-small cell lung tumours and examined the associations between p53 protein levels (corrected for total protein) and other clinico-pathologic variables, including post-surgical disease-free and overall survival. Contingency tables analysed by chi2 tests revealed no significant relationships between p53 status, defined by a median cut-off point, and patient gender, age, disease stage, histologic grade and type, lymph node extension, smoking history and administration of adjuvant chemotherapy or radiation. However, multivariate Cox proportional hazard regression analysis demonstrated a dose-response relationship between p53 concentration, expressed as a 4-level, quartile-divided variable, and increased risk of relapse (p = 0.010) and death (p = 0.016). Patients whose tumours contained p53 concentrations exceeding the median value had over 3-fold higher risk of relapse (p = 0.002) and death (p = 0.007) than those whose tumours had lower p53 concentrations. We also provide evidence suggesting that the impact of p53 on survival is greater in patients with squamous cell carcinoma than in those with adenocarcinoma. Although the latter finding needs confirmation, our results suggest that application of an immunoassay of p53 protein on non-small cell lung tumour extracts may identify patients at increased risk of unfavourable outcome.